> For the complete documentation index, see [llms.txt](https://whitepaper.gamegpt.gg/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://whitepaper.gamegpt.gg/08-the-gamegpt-game-dev-eval.md).

# The GameGPT game-dev eval

The Builder depends on the models behind it, and those models change constantly. Rather than pick by reputation, we are building our own evaluation for game development: a benchmark in the spirit of the coding benchmarks the industry already uses, but for making games.

Every leading model runs the same set of game-building tasks: game physics, gameplay mechanics and features, 2D and 3D asset creation, and 3D scene work. Each model is scored on each task. The results tell us which model to use for which job inside the Builder, and we intend to publish how the models compare. The eval is part of the GameGPT 2D roadmap.